Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <1616578270-7365-1-git-send-email-bingjingc@synology.com>
In-Reply-To: <1616578270-7365-1-git-send-email-bingjingc@synology.com>
Reply-To: fdmanana@gmail.com
From:   Filipe Manana <fdmanana@gmail.com>
Date:   Wed, 24 Mar 2021 11:30:12 +0000
Message-ID: <CAL3q7H7YGA0PFJp6J7vFvK0EPSixY7chnReS6Zbqa_9S2p_QRQ@mail.gmail.com>
Subject: Re: [PATCH] btrfs: fix a potential hole-punching failure
To:     bingjingc <bingjingc@synology.com>
Cc:     Josef Bacik <josef@toxicpanda.com>,
        David Sterba <dsterba@suse.com>,
        Qu Wenruo <quwenruo@cn.fujitsu.com>, Chris Mason <clm@fb.com>,
        linux-btrfs <linux-btrfs@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        cccheng@synology.com, Robbie Ko <robbieko@synology.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, Mar 24, 2021 at 11:15 AM bingjingc <bingjingc@synology.com> wrote:
>
> From: BingJing Chang <bingjingc@synology.com>
>
> In commit d77815461f04 ("btrfs: Avoid trucating page or punching hole in
> a already existed hole."), existed holes can be skipped by calling
> find_first_non_hole() to adjust *start and *len. However, if the given
> len is invalid and large, when an EXTENT_MAP_HOLE extent is found, the
> *len will not be set to zero because (em->start + em->len) is less than
> (*start + *len). Then the ret will be 1 but the *len will not be set to
> 0. The propagated non-zero ret will result in fallocate failure.
>
> In the while-loop of btrfs_replace_file_extents(), len is not updated
> every time before it calls find_first_non_hole(). That is, if the last
> file extent in the given hole-punching range has been dropped but
> btrfs_drop_extents() fails with -ENOSPC (btrfs_drop_extents() runs out
> of reserved space of the given transaction), the problem can happen.

This is not entirely clear. Dropping the last extent and still
returning ENOSPC is confusing.
I think you mean that it drops the last file extent item that does not
represent hole (disk_bytenr > 0), and after it there's only one file
extent item representing a hole (disk_bytenr =3D=3D 0).
It fails with -ENOSPC when attempting to drop the file extent item
representing the hole, after successfully dropping the non-hole file
extent item.
Is that it?

> After it calls find_first_non_hole(), the cur_offset will be adjusted
> to be larger than or equal to end. However, since the len is not set to
> zero. The break-loop condition (ret && !len) will not meet. After it
> leaves the while-loop, uncleared ret will result in fallocate failure.

Ok, fallocate will return 1, an unexpected return value.

>
> We're not able to construct a reproducible way to let
> btrfs_drop_extents() fails with -ENOSPC after it drops the last file
> extent but with remaining holes. However, it's quite easy to fix. We
> just need to update and check the len every time before we call
> find_first_non_hole(). To make the while loop more readable, we also
> pull the variable updates to the bottom of loop like this:
> while (cur_offset < end) {
>         ...
>         // update cur_offset & len
>         // advance cur_offset & len in hole-punching case if needed
> }
>
> Reported-by: Robbie Ko <robbieko@synology.com>
> Fixes: d77815461f04 ("btrfs: Avoid trucating page or punching hole in a
> already existed hole.")
> Reviewed-by: Robbie Ko <robbieko@synology.com>
> Reviewed-by: Chung-Chiang Cheng <cccheng@synology.com>
> Signed-off-by: BingJing Chang <bingjingc@synology.com>

Looks good.
Please just update that paragraph to be more clear about what is going on.

Thanks.

> ---
>  fs/btrfs/file.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
> index 0e155f0..dccb017 100644
> --- a/fs/btrfs/file.c
> +++ b/fs/btrfs/file.c
> @@ -2735,8 +2735,6 @@ int btrfs_replace_file_extents(struct inode *inode,=
 struct btrfs_path *path,
>                         extent_info->file_offset +=3D replace_len;
>                 }
>
> -               cur_offset =3D drop_args.drop_end;
> -
>                 ret =3D btrfs_update_inode(trans, root, BTRFS_I(inode));
>                 if (ret)
>                         break;
> @@ -2756,7 +2754,9 @@ int btrfs_replace_file_extents(struct inode *inode,=
 struct btrfs_path *path,
>                 BUG_ON(ret);    /* shouldn't happen */
>                 trans->block_rsv =3D rsv;
>
> -               if (!extent_info) {
> +               cur_offset =3D drop_args.drop_end;
> +               len =3D end - cur_offset;
> +               if (!extent_info && len) {
>                         ret =3D find_first_non_hole(BTRFS_I(inode), &cur_=
offset,
>                                                   &len);
>                         if (unlikely(ret < 0))
> --
> 2.7.4
>


--=20
Filipe David Manana,

=E2=80=9CWhether you think you can, or you think you can't =E2=80=94 you're=
 right.=E2=80=9D